Fair and Fast Convergence on Islands of Agreement in Multicriteria Group Decision Making by Logical Navigation
Résumé
Reasoning on multiple criteria is a key issue in group
decision to take into account the multidimensional nature of
real-world decision-making problems. In order to reduce the
induced information overload, in multicriteria decision
analysis, criteria are in general aggregated, in many cases
by a simple discriminant function of the form of a weighted
sum. It requires to, a priori and completely, elicit
preferences of decision makers. That can be quite arbitrary.
In everyday life, to reduce information overload people
often use a heuristic, called ``Take-the-best'': they
take criteria in a predefined order, the first criterion
which discriminates the alternatives at stake is used to
make the decision. Although useful, the heuristic can be
biased. This article proposes the Logical Multicriteria Sort process to support
multicriteria sorting within islands of agreement. It
therefore does not require a complete and consistent a
priori set of preferences, but rather supports groups to
quickly identify the criteria for which an agreement exists.
The process can be seen as a generalization of
Take-the-best. It also proposes to consider one criterion at
a time but once a criterion has been found discriminating it
is recorded, the process is iterated and relevant criteria
are logically combined. Hence, the biases of Take-the-best
are reduced. The process is supported by a GDSS, based on
Logical Information Systems, which gives instantaneous
feedbacks of each small decision and keeps tracks of all of
the decisions taken so far. The process is incremental,
each step involves low information load. It guarantees some
fairness because all considered alternatives are
systematically analyzed along the selected criteria. A
successful case study is reported.